Mobile Robot Path Planning Research Based On Genetic Algorithms

The robot is gradually coming to us，and has far-reaching impact in people’slives. Path planning technology，which is put forward higher request by complexenvironment and hymen’s expectations，is a key for the mobile robot to completethe task. The application of intelligent optimization algorithm in path planning hasits unique advantages，it can adapt to a changing environment，and more likeshuman beings making decisions. Genetic Algorithm(GA) doesn’t have too manyrestrictions for the optimization problem to be solved，and be good at solvingcomplex problems and nonlinear problems. It has nice flexibility，good implicitparallelism and global search capability to deal with problems. In mobile robot pathplanning，it has been widely appreciated，and made a series of achievements.However，the GA often has the shortcomings of precociousness and taking a longtime for convergence.This paper presents an approach to optimize GA using Case-Based Reasoning(CBR) method，in order to using it in the dynamic environment, the main researchare as follows.CBR stores a large number of former cases and a lot of experiences. Similarcases with the current issues for the GA are provided to improve the initial groupand accelerate the convergence process. In the evolutionary process，CBR can retainthe best individual to avoid the loss of the optimal solution for convergence to theglobal optimal solution.In a dynamic environment，it can induce GA search biases through periodicallyinjection relevant cases that matched with the environment，to achieve the optimalsolution and avoid premature convergence. So, it can adapt to the changingenvironment rapidly and find the viable path quickly.The evolutionary process is also the process for amending cases. In theprocess， we obtain optimal solution which is then stored into the case library，thelibrary is updated and guide future problem solving. Inspired by multistage decisionproblem， the re-optimization strategy is introduced to improve the acquireindividual，eliminating some of the constraints due to environmental modeling andindividual coding.In simple and complex static environment，GA is compared with CBR-GA insimulations， the fusion algorithm shows the superiority of the accelerated convergence rate and avoids premature effectively. The fusion algorithm is used forpath planning in dynamic environments，and obtains not bad results.